Sin categoría

QUIBIM_Curso IA Banner

CURSO PROPIO: Introducción a la Inteligencia Artificial aplicada a la Imagen Médica

Tenemos el placer de anunciar la primera edición de nuestro curso: Introducción a la Inteligencia Artificial aplicada a la Imagen Médica. 

Se trata del primer curso teórico-práctico centrado en dar conocer más a fondo algunos temas de máxima actualidad relacionados con la creación de algoritmos básicos de preproceso de las imágenes médicas y biomarcadores de imagen. Además, se practicará el desarrollo de algoritmos de inteligencia artificial basados en redes neuronales convolucionales aplicados a las imágenes médicas.

Dirigido a estudiantes de grado y máster de ingeniería biomédica, telecomunicaciones, informática, ciencia de datos, matemáticas, así como de otras carreras técnicas. Investigadores interesados en la imagen médica, la inteligencia artificial y los biomarcadores de imagen.

Puedes realizar la inscripción en las siguientes modalidades:

Recurso 4

PROGRAMA – Horario de 17 a 20 hrs.

DÍA 1 – Lunes 18 de noviembre de 2019 (3 horas)

  • Introducción al curso: la imagen médica desde Valencia al mundo.
  • Modalidades de adquisición:
  • No ionizantes (resonancia magnética y ultrasonidos)
  • Ionizantes (rayos X, tomografía computarizada y medicina nuclear) e imagen híbrida.

DÍA 2 – Martes 19 de noviembre de 2019 (3 horas)

  • Estándares en imagen médica: DICOM y otros.
  • Ejercicio-Notebook: Manipulación de formatos de imagen médica.
  • Repaso de conceptos.

DÍA 3 – Miércoles 20 de noviembre de 2019 (3 horas)

  • Procesamiento de imágenes médicas.
  • Biomarcadores de imagen: ¿Qué son?
  • Biomarcadores estructurales
  • Biomarcadores funcionales
  • Ejercicio-Notebook: Análisis de imágenes.

DÍA 4 – Jueves 21 de noviembre de 2019 (3 horas)

  • Introducción al Machine Learning
  • Ejercicio-Notebook: Inteligencia artificial aplicada: Clasificación con Redes Neuronales Convolucionales

FECHA: Del 18 de noviembre al 21 de noviembre de 2019

LUGAR: (COITCV) Col·legi Oficial d’Enginyers de Telecomunicació de la Comunitat Valenciana. Avinguda de Jacinto Benavente, 12. 46005, Valencia, España

¿Te interesa? Puedes realizar la inscripción en el siguiente enlace:

InscribemeMás info: PROGRAMA 

Abstract Hacking Codes. Hacker Computer Screen Closeup

Hacking Medical Images:
Fighting AI with AI

Recently, a new study led by Ben-Gurion University (Israel) in which AI was used to attack medical centers was published. The authors claimed they had developed a Deep Learning algorithm that was able to add or remove lung cancer from CT scans. The software was tested by attempting to mislead the diagnoses of three radiologists with 2, 5 and 7 years of experience, achieving an average success rate of 99.2% for cancer addition and 95.8% for cancer removal. These are alarming results, as the study proved that such attacks could be very harmful for medical institutions and their patients if somebody with bad intentions would use them. Therefore, designing countermeasures to guarantee the integrity of the medical data and protect health centers from this potential threat is a must for medical informatics departments. Below, we unveil what’s behind these AI cyberattacks and how they can be prevented and disarmed.

Imagen 1

Source: https://arxiv.org/pdf/1901.03597.pdf

The attack

The cyberattack that was proposed in the study was based in the use of Generative Adversarial Networks (GAN), a state-of-the-art Deep Learning technique introduced in 2014 by Ian Goodfellow, one of the most reputed AI researchers across the globe. The main idea behind this method lays in training two neural networks, called generator and discriminator, whose purpose is to defeat each other. On one hand, the objective of the generator network is to create fake data, that must be as realistic as possible. On the other hand, the goal of the discriminator network is to distinguish real data from the fake data created by the generator network. Both networks improve iteratively, as in each stage of the training the generator network is better at falsifying data and the discriminator is better at distinguishing false data, forcing each other to get better.

For the particular application of adding and removing lung cancer from CT studies there were two GAN models trained, one for the addition of lung cancer and another one for the removal and, therefore, there was a specialized generator network for the injection of fake tumors and another one specialized on removing them.

Source: https://skymind.ai/wiki/generative-adversarial-network-gan

Source: https://skymind.ai/wiki/generative-adversarial-network-gan

The defense

The first line of defense of medical centers should be handled by cybersecurity experts, designing communication networks robust to any kind of intrusion. But at QUIBIM we believe that even though cybersecurity is essential, there should be one more layer of protection against this kind of malware. We propose to defeat AI using AI if the hospital network is compromised. As explained before, GAN approaches consist of two networks. In the attack proposed in the study mentioned previously, the researchers developed a generator model able to misguide clinical decisions. Our defense approach is based on the opposite, that is, the design and implementation of an excellent discriminator network able to surpass the generator model and, therefore, being able to detect the artificially generated images. This network purpose would be to learn to look beyond what a human eye can detect, which would make a perfect barrier against generative attacks.

Some final words

In conclusion, Ben-Gurion University publication helped to highlight that even though AI unlocked many new opportunities and allowed major breakthroughs in the medical sector, it also favored the development of potentially harmful applications. Hence, there is a need for medical centers to build a strong AI strategy, not only to improve patient care, but to be prepared and protected against AI malware. Because, as all revolutionary technologies, AI can be used for the best, but for the worst as well.

At QUIBIM, our mission is to improve humans’ health by applying advanced and innovative image processing techniques to radiological images, and we guarantee that we will make use of our extended expertise and advanced algorithms to be prepared against any malignant software that aims to mislead the results of our analysis pipelines or our customers’ clinical outcomes.

ESOR_QUIBIM Course

QUIBIM provides the platform for the GALEN Advanced Course organized by European School of Radiology

  • QUIBIM provided the software for the hands-on workshop activities, allowing a much more dynamic and interactive case discussion

QUIBIM was honored with the chance to participate in the last edition of the “GALEN Advanced Course on Oncologic Imaging of the Abdomen”. The event was organized by the European School of Radiology (ESOR), the educational initiative of the European Society of Radiology (ESR).

This course was aimed at senior residents, board-certified radiologists and fellows interested in abdominal oncologic imaging and focused on the application of the latest technical advancements and the new European guidelines for imaging.

ESOR_QUIBIM PrecisionQUIBIM provided the software for the hands-on workshop activities, allowing a much more dynamic and interactive case discussion. Based on our QUIBIM Precision® cloud Platform, this tool provides a powerful framework for the creation of new users, the uploading of imaging studies and relevant documentation and for the administration of the course.

Furthermore, the platform has been developed to provide lecturers with convenient features to share studies with students, visualize and edit them using our zero-footprint embedded DICOM Web Viewer and, most important, analyze the studies with any of the imaging biomarker plugins available at QUIBIM Precision®.

This event represents a great milestone for QUIBIM, as ESOR, with over 19.000 participants in more than 250 ESOR courses, has become the major provider of complementary radiological education in Europe and worldwide.

ESOR_QUIBIM

QUIBIM_FEDER2-IVACE

QUIBIM recibe la ayuda del IVACE – Proyectos de I+D de PYME (PIDI-CV)

QUIBIM ha obtenido financiación del Institut Valencià de Competitivitat Empresarial (IVACE) dentro del programa “Proyectos de I+D de PYME (PIDI-CV)” para la realización del proyecto “DESARROLLO DE ALGORITMOS DE INTELIGENCIA ARTIFICIAL PARA LA DETECCIÓN AUTOMATIZADA DE VÉRTEBRAS A PARTIR DE IMÁGENES DE TC EN OSTEOPOROSIS” con número de Expediente: IMITDA/2017/120.

OBJETIVOS Y RESULTADOS DEL PROYECTO

El presente proyecto tiene el objetivo de desarrollar nuevas técnicas de análisis de imagen y algoritmos de Inteligencia Artificial (Machine Learning y Deep Learning) aplicados  a la caracterización de la columna vertebral a través de la segmentación automática de vértebras en imágenes de TC (tomografía computarizada), que permita el soporte al radiodiagnóstico en pacientes con osteoporosis. Para ello, se crea una nueva herramienta software de soporte que permita detectar e identificar automáticamente las vértebras en una imagen de TC. Una vez integrado este nuevo desarrollo como un nuevo módulo en nuestra plataforma Quibim Precission, estaremos en disposición de  caracterizar la microarquitectura ósea de cada una de las vértebras, realizar una evaluación de la misma y ofrecer un radiodiagnóstico avanzado capaz de aportar mayor información sobre la enfermedad ósea del paciente. 

La propuesta de valor para QUIBIM es contribuir con el presente proyecto a completar sus líneas de I+D, en concreto su línea  musculoesquelética. Así este proyecto ha supuesto la creación de un sistema especializado en la caracterización automática de la microarquitectura ósea vertebral, ofreciendo al radiólogo información cuantitativa sobre ésta, para mejorar la evaluación de tratamientos médicos y agilizar el control y seguimiento de pacientes con osteoporosis.

QUIBIM_chest_xray_classifier_logo3

New Chest X-Ray Classification Tool

Despite the technological evolution of imaging modalities like CT, US and MRI, conventional X-ray remains the most performed examination in radiology departments, and remains a fundamental tool for anatomical analysis in the detection and diagnosis of respiratory diseases and bone tissue alterations. However, radiology departments have limitations in reporting the X-Rays due to the limited resources available (link).

QUIBIM has developed a Chest X-Ray Classification Tool that offers a solution to this problem which can help radiology departments become even more efficient. This classifier developed in collaboration with Hospital Universitario y Politécnico La Fe, estimates the probability of chest X-Rays of having a pathology using Artificial Intelligence.

How does it work?

This tool makes use of fourteen Convolutional Neural Networks trained with a database of more than 100,000 images (NIH ChestXray14 dataset) to estimate the probability of presence of the following pathologies in chest X-Rays: atelectasis, cardiomegaly, effusion, infiltration, mass, nodule, pneumonia, pneumothorax, consolidation, edema, emphysema, fibrosis, pleural thickening and hernia. Afterwards, the probabilities are used by a Fully Connected Neural Network to get the final probability of the X-Ray of being abnormal.

XRAY chest

Because of this AI methodology the classifier understands the visual patterns that are most indicative of the different pathologies using the knowledge extracted from the large dataset of radiographs used to train the networks. QUIBIM’s Chest X-Ray Classification Tool is able to learn further using new images, which means that this system is continually improving and evolving with time.

 What information does it provide?

Once all the quantifications are performed this tool provides an intuitive Structured Report with the patient’s information, the abnormal probability of the radiographs analyzed and the representation of the findings. In addition, if the image is classified as abnormal, the report shows the three pathologies that are more likely to be found and a heatmap that highlights the most abnormal regions. This report is designed to be very user friendly, to assist the user in understanding the tool’s findings at a glance in order to make its usage highly efficient.

Why is this new tool so useful?

Using this technology it is possible to prioritize unreported, potentially pathological radiographs which allows radiologists to focus their efforts on studies that are more likely to have pathologies and thereby become more efficient. This tool essentially ensures that pathological findings, which could have been unreported due to the heavy workload of radiology departments, are correctly reported.

QUIBIM’s goal is to provide the radiology departments with an optimal solution for automatic reading of X-Rays without interfering in the  workflow of the department.

QUIBIM’S Chest X-Ray Analysis Tool is already available at QUIBIM Precision® Depending on the needs of the department, this tool is accessible through the cloud with just a few clicks  or  it can be fully integrated in the radiology department’s workflow as a local solution for  seamless interpretation of  chest X-Rays.

boton try quibim

ESGAR Conference CCD Dublin Ireland 2018.

Co-Founder Prof. Luis Martí-Bonmatí receives ESGAR Gold Medal

During the 29th Annual Meeting and Postgraduate Course of ESGAR (European Society of Gastrointestinal and Abdominal Radiology) held in Dublin, our co-founder, Prof. Luis Marti-Bonmati received the ESGAR Gold Medal, the society’s highest award for outstanding contributions to the scientific community.

All QUIBIM team is proud and honoured. We congratulate Prof. Martí-Bonmatí on this achievement.

ESGAR Conference CCD Dublin Ireland 2018.

ESGAR Conference CCD Dublin Ireland 2018.

Photo credit: Roger Kenny Photography

Photo legend (from left to right): Steve Halligan (ESGAR President), Luis Martí-Bonmatí (Gold Medallist), Celso Matos (ESGAR Past President), Helen Fenlon (ESGAR Meeting President 2018).

Link to Insights into Imaging post: https://www.i3-journal.org/news/esgar-goldmedal/

 

QUIBIM Imaging Biomarkers made transparent

QUIBIM, AI imaging disruption to showcase the value of Precision

We are in the era of Precision Medicine, and so is Radiology. Nowadays, main imaging modalities like X-ray, computed tomography (CT), magnetic resonance (MR), positron emission tomography (PET) and hybrid machines, among others, have become measurement instruments. Images are not only pictures anymore, but thanks to the application of computational analysis and artificial intelligence, they are data, as you can learn in this excellent manuscript in Radiology.

Nowadays, when radiologists perform measurements of different organs, tissues and lesion properties using a workstation, they are used to get a number (i.e. lesion volume or perfusion). If they take the same images and they get to analyze them in a workstation from another vendor (not straightforward), it is pretty sure that they will obtain different results. This issue has introduced a sense of lack of standardization and homogenization in the quantitative medical imaging field.

I like to say that value is to trust in the product, and we have decided to be the first company in the world to open the validation process and tests results of our imaging biomarkers. Every time we buy a measurement device for daily life purposes (i.e. thermometer) we know the degree of uncertainty, why wouldn’t we do the same in AI algorithms and quantitative imaging?

We are proud to make this announcement at ECR 2018: Now it is possible to see the precision, accuracy and clinical evaluation results of our imaging biomarkers. We provide the precision (through Coefficient of Variation, CoV) and accuracy (through relative error, e) values through the publication of QUIBIM Technical Datasheets that you can find in the resources section of our webpage.

With this strategy QUIBIM is going a step further by being the first multi-vendor, web-based and real precision Medicine company of the medical imaging & AI market.

Concerned by the accuracy of your measurements? Let’s work together.

QUIBIM RECIBE LA AYUDA DEL IVACE-INTERNACIONALIZACION 2017

QUIBIM es un proyecto empresarial de alto impacto social y sanitario, que extrae información cuantitativa de las imágenes médicas radiológicas, mediante técnicas innovadoras y avanzadas de procesado computacional, con el objetivo de mejorar los procesos de diagnóstico de enfermedades con alta incidencia y evaluar adecuadamente los cambios que producen los tratamientos farmacológicos en el organismo.

025-FEDER2-declaracion14-20

Durante 2017 el proyecto de internacionalización QUIBIM ha recibido la ayuda IVACE – “ACCIONES DE PROMOCIÓN EN EL EXTERIOR QUIBIM 2017” (ITAPIN/2017/447) con el apoyo del Fondo Europeo de Desarrollo Regional (FEDER) por un importe de 4.405,16€

IVACE_QUIBIM
QUIBIM_MIUC

MIUC
the new toolkit of QUIBIM Precision® platform to beat traditional workstations

Quibim has implemented a new toolkit named MIUC (Medical Imaging Universal Connector) to close the gap between hospital IT systems and the Cloud. Whereas the Cloud satisfies the processing requirements, Quibim Precision® handle the functionalities related to communications and management of DICOM objects among the hospitals and radiological centers.

Quibim Precision® allows users in hospitals and radiology departments to have a seamless integration of imaging biomarkers analysis within the radiological workflow, due to the MIUC capabilities combined by the Quibim Precision® Cloud computing environment and the interoperability features implemented in our system. Both image upload and data retrieval are fully automated and users only need to access the PACS when they are notified that a new biomarker report is available.

MIUC is placed inside the hospitals and clinics and it is responsible to establish all the required communications between the PACS and Quibim Precision®. During the analysis of imaging biomarkers, the study is anonymized, sent to the Cloud and analyzed. The final result is a one-page report which is sent back to the MIUC or can be directly visualized in the Quibim Precision® web interface. Furthermore, in clinical environments, the report is converted into DICOM objects and stored in the PACS as a new series within the original study. To identify the original study, the MIUC implements backward traceability in the client side to reidentify the anonymized studies.

Our platform is intended to be used by radiologists, either from a clinical environment, thanks to the MIUC, or as a final user using the web interface. In the clinical environment scenario, radiologists using Quibim Precision® do not have to worry about where the study or the report is. Instead, these issues are transparent to the user, who do not have to perform any action to launch a biomarker process, given that the MIUC rule engine does such work for them. The user will be notified by email when a new biomarker report is ready (and available in both the PACS and the Quibim Precision® web interface).

Nowadays, our Quibim Precision® platform is compliant with the DICOM standard at both communication level and data management and formatting level. Specifically, our platform receives imaging studies from hospitals, radiological centers or pharma companies. Then, the system analyzes the study and obtains quantitative measures, which are stored in a quantitative database and structured in a one-page report on a per-patient basis. Finally, this report is returned back as a result.  Quibim Precision® allows annotating biomarker reports using terms from RadLex and MeSH, enhancing the interoperability of its biomarker reports with other health information systems. In fact, the imaging platform is seamlessly integrated with the hospital PACS, being able to query and retrieve medical studies, processing them and storing the resulting biomarker reports as DICOM objects in the hospital PACS. On the other hand, the processing stage is performed on the Cloud, taking advantage of its benefits: high-performance computing and real-time hardware scalability on demand.

But, what has changed?

In previous updates of our platform, we improved the performance, capabilities and user settings view. With this new suite software QUIBIM Precision®- MIUC does query/retrieve the PACS, anonymizes the PACS responses and forwards them to Quibim Precision® in the Cloud. Furthermore, the MIUC leads our solution to a higher level of automation, given that it monitors the PACS querying for incoming studies. Once a new study reaches the PACS, the MIUC analyzes its header and determines whether a new biomarker analysis must be launched or not, depending on some DICOM elements in the study like imaging modalities, study description, series description or body part among others. For each biomarker analysis available in Quibim Precision®, there is a predefined set of rules that establishes which studies are susceptible to be processed by each analysis method. An incoming study matches a given analysis method whenever it fulfils the predefined set of rules for such analysis method. When this happens, the MIUC automatically sends the study to the Quibim Precision® Cloud processing platform, where it will be processed by the matching biomarker analysis pipeline. Once processed, a biomarker report is generated with the results and sent back to the MIUC. Finally, the MIUC stores the report in the PACS as a DICOM object, making the report available for the specialist who requested it. This way, the Cloud platform remains centralized and, at the same time, fully integrated with the hospital IT systems.

With the arrival of MIUC toolkit the need for conventional workstations with expensive licenses in radiology departments completely disappears. As in other business areas that are evolving from product to service, the Quibim image analysis technology was designed to be offered as the service that puts disruptive image analysis solutions at your fingertips.

QUIBIM at BIO 2017

QUIBIM in BIO 2017

Our company was present again this year in the incredibly huge BIO International Convention in San Diego, California, from June 19th to 22nd. Our registration included both a booth at the Spanish pavilion and the access to the One-to-One partnering meetings.

We had 15 planned meetings and many other new contacts thanks to the interaction at our exhibitor space with agents interested in QUIBIM business model. We made several demonstrations of QUIBIM Precision platform and image analysis capabilities in clinical trials. From all the contacts and meetings performed at BIO, I wanted to point out the classification we found according to their profile:

  • Scientific parks and incubators (30%)
  • Investors in Life Sciences (20%)
  • CRO’s and Pharma companies (50%)

From our experience last year in San Francisco, in this edition there has been a higher interest from scientific parks and incubators beyond Boston and Silicon Valley to attract companies to their facilities, showing the benefits of establishing the companies in specific locations, specially in the different states of US. The number of investors stayed similar, but we have been an increasing interest in the field of Medical Devices. Regarding CRO’s and Pharma companies, most of them are progressively considering medical imaging in their clinical trials, and the best, considering us for their solutions. We are so proud to cover those unmet needs on advanced image analysis services for Clinical Trials, allowing pharma companies, CRO’s and Principal Investigators to follow-up in real time their study. In fact, one of the main trends at BIO this year was how data processing will change the way new drugs are developed and launched into market.

QUIBIM CEO (Angel Alberich-Bayarri) & Booth at BIO 2017

QUIBIM CEO (Angel Alberich-Bayarri) & Booth at BIO 2017

 

We were so glad to have this exhibitor space at the Spanish pavilion, and compared to previous editions, it was also the first time that the Valencia region had a dedicated area inside it (similar to Biocat from Catalonia and Biobasque from Basque Country). The Valencia area was organised by IVACE (Instituto Valenciano para la Competitividad Empresarial), and the organism was represented by Mrs. Mónica Payá (representative for foreign investment of IVACE). The Principe Felipe Research Centre (CIPF), was also represented by Oscar David Sánchez (Projects and Technology Transfer Manager).

Valencia region representatives at BIO 2017 in San Diego, Angel Alberich (QUIBIM), Mónica Payá (IVACE), Daniel Calvo (BIOPOLIS), Marisol Quintero (Biooncotech)

Valencia region representatives at BIO 2017 in San Diego, Angel Alberich (QUIBIM), Mónica Payá (IVACE), Daniel Calvo (BIOPOLIS), Marisol Quintero (Biooncotech)

 

Valencia region representatives at BIO 2017 in San Diego, Óscar David Sánchez (CIPF), Mónica Payá (IVACE), Daniel Calvo (BIOPOLIS), Angel Alberich-Bayarri (QUIBIM)

Valencia region representatives at BIO 2017 in San Diego, Óscar David Sánchez (CIPF), Mónica Payá (IVACE), Daniel Calvo (BIOPOLIS), Angel Alberich-Bayarri (QUIBIM)

 

All the days at BIO were so productive that there is a significant work to be done at home, contacting back with the people we met and following up these new relationships.

Obviously not everything is work and there is also some spare time for entertainment at BIO, in the following picture, a rock band playing at the middle of Gaslamp quarter in San Diego. The streets were closed to welcome BIO 2017 participants in a nice evening with food, drink and music, a nice experience!

Band performing at BIO 2017 in middle of Gaslamp quarter

Band performing at BIO 2017 in middle of Gaslamp quarter